A Geomorphology Based Reconstruction of Ice Volume Distribution at the Last Glacial Maximum Across the Southern Alps of New Zealand James, William H

Aberystwyth University

A geomorphology based reconstruction of ice volume distribution at the Last Glacial Maximum across the Southern Alps of New Zealand James, William H. M.; Carrivick, Jonathan L.; Quincey, Duncan Joseph; Glasser, Neil

Published in: Quaternary Science Reviews DOI: 10.1016/j.quascirev.2019.06.035 Publication date: 2019 Citation for published version (APA): James, W. H. M., Carrivick, J. L., Quincey, D. J., & Glasser, N. (2019). A geomorphology based reconstruction of ice volume distribution at the Last Glacial Maximum across the Southern Alps of New Zealand. Quaternary Science Reviews, 219, 20-35. https://doi.org/10.1016/j.quascirev.2019.06.035

Document License CC BY

General rights Copyright and moral rights for the publications made accessible in the Aberystwyth Research Portal (the Institutional Repository) are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the Aberystwyth Research Portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the Aberystwyth Research Portal

Take down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim. tel: +44 1970 62 2400 email: [email protected]

Download date: 07. Oct. 2021 Quaternary Science Reviews 219 (2019) 20e35

Contents lists available at ScienceDirect

Quaternary Science Reviews

journal homepage: www.elsevier.com/locate/quascirev

A geomorphology based reconstruction of ice volume distribution at the Last Glacial Maximum across the Southern Alps of New Zealand

* William H.M. James a, , Jonathan L. Carrivick b, Duncan J. Quincey b, Neil F. Glasser c a School of Geography and Leeds Institute for Data Analytics, University of Leeds, Leeds, West Yorkshire, LS2 9JT, UK b School of Geography and water@leeds, University of Leeds, Leeds, West Yorkshire, LS2 9JT, UK c Department of Geography and Earth Sciences, Aberystwyth University, Wales, SY23 3DB, UK article info abstract

Article history: We present a 3D reconstruction of ice thickness distribution across the New Zealand Southern Alps at the Received 11 February 2019 Last Glacial Maximum (LGM, c. 30e18 ka). To achieve this, we used a perfect plasticity model which Received in revised form could easily be applied to other regions, hereafter termed REVOLTA (Reconstruction of Volume and 26 June 2019 Topography Automation). REVOLTA is driven by a Digital Elevation Model (DEM), which was modified to Accepted 28 June 2019 best represent LGM bed topography. Specifically, we removed contemporary ice, integrated offshore bathymetry and removed contemporary lakes. A review of valley in-fill sediments, uplift and denudation was also undertaken. Down-valley ice extents were constrained to an updated geo-database of LGM ice Keywords: fi Geomorphology limits, whilst the model was tuned to best- t known vertical limits from geomorphological and 3 Glacial geochronological dating studies. We estimate a total LGM ice volume of 6,800 km , characterised pre- Glaciation dominantly by valley style glaciation but with an ice cap across Fiordland. With a contemporary ice Glaciology volume of approximately 50 km3, this represents a loss of 99.25% since the LGM. Using the newly created Quaternary ice surface, equilibrium line altitudes (ELAs) for each glacier were reconstructed, revealing an average Southern pacific ELA depression of approximately 950 m from present. Analysis of the spatial variation of glacier-specific ELAs and their depression relative to today shows that whilst an east-west ELA gradient existed during the LGM it was less pronounced than at present. The reduced ELA gradient is attributed to an overall weakening of westerlies, a conclusion consistent with those derived from the latest independent climate models. © 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

1. Introduction conditions. Glacial reconstruction in New Zealand presents significant The New Zealand Southern Alps is a key site for glacial recon- challenges as the extreme relief and high precipitation of parts of struction studies and is one of only three mid-latitude southern the Southern Alps gives rise to a complex yet large-scale system hemisphere locations at which extensive Pleistocene glaciation (Chinn et al., 2014). To date, the work by Golledge et al. (2012) is the occurred. Furthermore, the geomorphological record of the Last only attempt at a regional reconstruction of LGM ice thickness, and Glacial Maximum (LGM) in New Zealand has excellent spatial and that was achieved using the Parallel Ice Sheet Model (PISM). They temporal coverage (Barrell et al., 2011) and has long been recog- used a Digital Elevation Model (DEM) and climate parameter inputs nized for its potential for improving understanding of former cli- (spatially uniform scaling of contemporary conditions) to best fit matic conditions via glacier geometry reconstruction (e.g. Porter, the areal ice limits delimited by Barrell (2011) (Fig. 1). Some rela- 1975b). LGM ice thickness has previously been reconstructed for tively large (tens of kilometers) discrepancies must be noted individual catchments in New Zealand (e.g. Putnam et al., 2013) and however, where the PISM was unable to recreate known down- at the regional scale (Golledge et al., 2012), although uncertainty valley limits derived from field-based observations and indepen- still exists regarding the extent, timing and associated climatic dently dated landforms. Specifically, in comparison to the mapping of Barrell (2011), Golledge et al. (2012) underestimated ice extent in the eastern outlets including the Rakaia and Waimakariri (Fig. 1).

* As an alternative to top-down climate models, a number of Corresponding author. ‘ ’ E-mail address: [email protected] (W.H.M. James). bottom-up, topography based techniques have been developed for https://doi.org/10.1016/j.quascirev.2019.06.035 0277-3791/© 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35 21

Fig. 1. Location map of regions referred to in this study and previous delineations of LGM ice extent. Note the mismatch between extents estimated by Barrell (2011) and Golledge et al. (2012). Inset: Modelled ice thickness and extent by Golledge et al. (2012) and extent mapped by Barrell (2011) for the Rakaia Valley. estimating contemporary ice thickness (Linsbauer et al., 2012; applied a topography driven model to produce a new estimate of James and Carrivick, 2016) and palaeo-ice thickness (Pellitero et al., distributed LGM ice thickness. Our approach is similar to that 2016). With the Southern Alps presenting excellent input datasets described by Pellitero et al. (2016) and in effect a reverse of the for LGM ice thickness reconstruction in this manner (high resolu- VOLTA contemporary ice thickness model developed by James and tion DEMs and well constrained down-valley ice limits), and with Carrivick (2016). As such the model is hereafter referred to as climate driven models currently unable to resolve known ice limits REVOLTA (Reconstruction of Volume and Topography Automation). in some catchments (i.e. Golledge et al., 2012), we developed and Using this approach, the aim of this study is to reconstruct the 22 W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35

LGM distributed ice thickness across the Southern Alps of New bathymetric dataset (Mitchell et al., 2012) and these data were Zealand to help improve our knowledge of glaciological and asso- appended to the existing DEM. To create a seamless dataset, the ciated climatic conditions during the LGM in the Southern Hemi- bathymetry was resampled to the resolution of the land surface sphere. Being far from the centres of Northern Hemisphere (8 m) and merged. glaciation, research in New Zealand provides the potential to examine the climatic effects of changes in Northern Hemisphere ice 2.1.3. Contemporary lakes sheets on Southern Hemisphere climate, as well as events of origin Whilst palaeoglaciers extended past the contemporary coast- in the Southern Hemisphere (Shulmeister et al., 2019). line, they also occupied areas where contemporary lakes now exist, for which standard DEMs represent the water surface level and not 2. Methods: a framework for reconstructing distributed ice the underlying bathymetry, thus obscuring the former glacier bed thickness topography. Substantial lakes inboard of LGM moraine sequences of the Southern Alps are over 450 m deep (e.g. Irwin, 1980), high- REVOLTA initially calculates the surface profiles of former gla- lighting the importance of considering their bathymetry. Many of ciers along a network of centrelines using an exact solution of a these lakes developed at the end of the LGM and were in contact ‘perfect plasticity’ approach. Ice thickness values are iteratively with a calving ice margin (Sutherland et al., 2019). calculated along each centreline, starting at the former terminus To produce a DEM more representative of LGM conditions, the position. A ‘nearest neighbour’ routine is used in conjunction with water depth of major lakes within the model domain was sub- the DEM to convert the centreline thickness points into a 3D tracted from the initial DEM. Each map was georeferenced, with distributed thickness. REVOLTA expands on the 2D only model water depth contours ranging in interval between 5 m and 100 m presented by Benn and Hulton (2010) and is similar in nature to digitized. Bathymetry was adjusted to the lake level of the DEM Pellitero et al. (2016). A schematic diagram of the REVOLTA before a gridded surface was interpolated using the hydrologically framework is shown in Fig. 2 and major processing steps are correct ANUDEM algorithm (Hutchinson, 1989). These data were described in the following sections. subsequently merged with the DEM, helping to produce a surface more representative of LGM bed conditions. 2.1. Constructing a DEM to represent the LGM Whilst sediment deposition has occurred in these lakes since the LGM, there is insufficient data to remove this from the DEM. Of A digital elevation model (DEM) forms the primary input of the limited data which does exist, cores from Lake Ohau found REVOLTA. Whilst contemporary DEMs are often used to represent sediment thicknesses (deposited since lake formation at the end of former glacier beds (Trommelen and Ross, 2010), the most accurate the LGM) of 82 m and 43 m beneath 101 m and 68 m of water results will be achieved if the DEM is representative of the period of respectively (Levy et al., 2018). This shows that whilst the presence reconstruction (i.e. the LGM). To achieve this, we used a contem- of such sediment needs to be acknowledged, water depth alone porary 8m resolution DEM (Geographix, 2012), to which we applied accounts for the majority of the difference between the LGM bed a number of corrections: (i) removing contemporary ice thickness, and the contemporary DEM surface. By removing the water depth, (ii) integrating offshore bathymetry, (iii) removing modern-day our study accounts for over 60% of the difference between the LGM lakes that fill the remnants of troughs evacuated by LGM ice. The bed and the contemporary DEM, something which has not been impact of uplift, denudation and postglacial sediment valley infill considered by previous studies (e.g. Golledge et al., 2012). was also considered. 2.1.4. Uplift and denudation 2.1.1. Removal of contemporary ice from the DEM Uplift and denudation are processes that modify the landscape, The Southern Alps currently supports around 3,000 glaciers meaning their influence on the DEM needs to be considered. Whilst covering an area of approximately 1,200 km2 (Chinn, 2001). these changes are commonly too small to introduce significant Contemporary ice was removed from the DEM to estimate a surface error (Finlayson, 2013) and are thus commonly not included in ice which better represents LGM bed conditions. This removal was thickness modelling studies (Golledge et al., 2012; McKinnon et al., achieved using the distributed ice thickness results of the VOLTA 2012), here we assess how robust their omission is in terms of model (James and Carrivick, 2016) to compute modern day ice impacting ice volume of the Southern Alps. thickness (e.g. as for Southern South America by Carrivick et al. Post-LGM isostatic rebound in New Zealand is considered to (2016) and as for the entire Antarctic Peninsula by Carrivick et al. have been minimal (Pickrill et al., 1992), with Mathews (1967) (2019)). VOLTA was run on all contemporary glaciers >1km2 and suggesting a maximum of 30 m rebound. However, tectonic uplift resulted in a total (contemporary) ice volume of 40.68 km3 with a rates in the Southern Alps are high, with the Pacific plate being maximum thickness of over 550 m. Full details of the VOLTA model compressed into the Australian plate (Coates, 2002). Whilst long can be found in James and Carrivick (2016), whilst details of the term Cenozoic uplift rates of 0.1e0.3 mm yr 1 were estimated by contemporary ice thickness distribution of the Southern Alps can fission track thermochronology (Tippett and Kamp, 1995), rates be found in James (2016). during the Pleistocene and Holocene are thought to have been greater (Adams, 1980). The Paringa River site (location shown in 2.1.2. Appending an offshore DEM Fig. 1) provides one of the few robust constraints, with Pleistocene Sea level during the LGM was approximately 125 m below cur- to Holocene rates calculated at 8 ± 1mmyr 1 (Norris and Cooper, rent levels (Milne et al., 2005), with palaeoglaciers on the West 2001). However, this rate is not thought to be representative of Coast of New Zealand inferred to extend into these currently sub- the Southern Alps as a whole as the uplift pattern was highly marine environments. Specifically, analysis of areal LGM limits spatially variable, as evidenced by tilted lake shorelines (Adams, delineated by Barrell et al. (2011) revealed that 2.5% of a total LGM 1980). glaciated area occurred beyond the present coastline and conse- Since AD 2000, GPS measurements have been collected along a quently outside the limits of the standard DEM. Therefore, an transect running broadly perpendicular to the Alpine Fault, alternative topographical source was required for these presently- providing further information on the distribution of uplift rates offshore areas. For this purpose, we used the National Institute of across the Southern Alps. Results show a maximum average vertical Water and Atmospheric Research (NIWA) 250 m resolution rate of 5.4 mm yr 1 for the transect (Beavan et al., 2007). However, W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35 23

Fig. 2. Schematic overview of the workflow developed for estimating palaeo ice thickness (REVOLTA). and critically, rates reduce rapidly with distance from the Alpine Mount Cook) is 3,724 m high, the total amount of uplift during Fault, with measurements 29 km from the fault recording an uplift orogeny is estimated to be approximately 15e20 km (Kamp et al., rate of just 1.8 ± 1mmyr 1, whilst no significant vertical rise was 1989). Tippett and Kamp (1995) provide one of the few quantifi- measured at 68 km (Beavan et al., 2007). As these GPS measure- cations of denudation, using fission track thermochronology to ments record current rates, it is acknowledged that they are not estimate long term Cenozoic denudation rates ranging from necessarily applicable to longer (glacial) timescales. Nonetheless, ~2.5e0.5 mm yr 1, decreasing with distance from the Alpine Fault. with contemporary uplift rates reducing exponentially with dis- This is supplemented by contemporary river load measurements by tance from the Alpine Fault and no significant rise recorded at Adams (1980) who estimated the total river load of the Southern 68 km from the fault, and in lieu of any other geological data per- Alps to be 700 ± 200 109 kg yr 1 whilst basin averaged rates for taining to long-term and spatially-distributed uplift rates, it is likely the Nelson/Tasman region are estimated to be between 112 and that much of the model domain for an LGM ice reconstruction 298 t km 2 yr 1 (Burdis, 2014). Whilst providing a useful insight would fall within relatively low uplift areas. This interpretation is into denudation rates, these estimates are not sufficient to consistent with the inferred uplift pattern of Adams (1980), with construct a regional distribution required for DEM modification. the majority of the LGM ice extent occupying low uplift regions. In summary, whilst it is acknowledged that uplift and denuda- Importantly, denudation effectively counteracts uplift, with long tion have modified the geomorphological expression of the land- term rates in the Southern Alps approximately equal (Adams, 1980). scape, the approximate counterbalance between the two processes For example, whilst the highest peak in the Southern Alps (Aoraki, limits their impact between the land surface at the LGM and the 24 W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35 contemporary DEM. Furthermore, with rates of both uplift and without removing post LGM sediment. denudation decreasing with distance from the Alpine Fault, much of our model domain falls within relatively low uplift/denudation 2.1.5.1. Distributed post-LGM sediment thickness of the Pukaki Valley. areas. With the New Zealand land mass largely in its present As discussed in section 2.1.5, regional modelling was performed configuration by the mid-Pleistocene (Barrell, 2011), we therefore without removal of any post-LGM sediment as there is insufficient made no further modifications to the DEM. These issues of tectonic data to generate a regional distribution. However, modelling by deformation and denudation will become increasingly prevalent McKinnon et al. (2012) provides sufficient information to estimate for older glaciations, so care must be taken if seeking to reconstruct post-LGM sediment thickness for the Pukaki Valley and conse- earlier glacial episodes. quently allowed us to test the sensitivity of REVOLTA at this location. Specifically, McKinnon et al. (2012) found that a mass-flux 2.1.5. Postglacial sediment valley infill balance model (as described by Anderson et al. (2004)) produced Large volumes of sediment accumulated in valley floors is a the most reliable LGM bed profile and associated post-LGM sedi- commonly noted feature of the Southern Alps (McKinnon et al., ment thicknesses. The reader is directed to McKinnon et al. (2012) 2012). This process of landscape modification has previously been and Anderson et al. (2004) for a full discussion of the model choice recognized as a potential source of uncertainty when reconstruct- and its mechanics. ing palaeoglaciers in New Zealand (Golledge et al., 2012) and needs Spatially distributed post-LGM sediment thickness was esti- to be considered when using a DEM to represent a former glacier mated by firstly identifying the areal extent of sediment in-fill and beds. subsequently interpolating the sediment thickness in this region It was initially postulated that the majority, or entirety of valley using the transect data of McKinnon et al. (2012). The areal extent fill in the New Zealand Southern Alps was deposited post-LGM was identified using a modified version of the algorithm presented (Suggate, 1965; Adams, 1980) simply due to the ‘absence of any by Straumann and Korup (2009), using a GIS derived stream conflicting evidence’ Adams (1980, p. 77), with some contemporary network as ‘seed cells’ and iteratively assessing the relative eleva- research supporting the hypothesis that glaciers did indeed erode tion of neighboring cells. Specifically, we set a gradient threshold of to bedrock at the LGM (Thomas, 2018). However, this notion has 1.5 between neighboring cells as this was found to best match the been challenged for the down valley reaches of LGM glaciers in the visual geomorphological evidence. The mathematics of the algo- Southern Alps, with Rother et al. (2010) suggesting that recent rithm can be found in Straumann and Korup (2009) and for our glaciations in the Hope Valley (see Fig. 1 for location) were not purposes we coded the process in the Python programming lan- powerful enough to cause extensive erosion and overrode pre-LGM guage. Once the areal extent of sediment infill was delineated, we sediments rather than removing them. At this location, infrared interpolated the post LGM thickness within this region using the stimulated luminescence (IRSL) dating and sediment analysis found mass flux balance bed profile generated by McKinnon et al. (2012). the majority of material was deposited during the pre-LGM period Interpolation was achieved using the ANUDEM algorithm as this is of 95.7 to 32.1 ka with only approximately 30 m of sediment designed specifically for topographical applications (Hutchinson, accumulation occurring post LGM, overriding 200 m of previous 1989). deposits (Rother et al., 2010). This pattern is consistent with seismic investigation in the Franz Josef Valley by Alexander et al. (2014), 2.2. Areal extent of LGM glaciers suggesting that the LGM bed was well above the bedrock. A modelling study by McKinnon et al. (2012) also supports the Alongside the DEM, areal ice extents are required as a model concept of LGM glaciers overriding pre-existing sediment, using a input for REVOLTA, with the downstream point of each centreline variety of models to simulate the LGM bed of the Pukaki glacier. The defining the starting point for each glacier reconstruction. The notion of pre-LGM sediment surviving in glaciated valleys is also latest revision of LGM ice extent mapping was produced by Barrell consistent with studies outside of New Zealand, with pre LGM (2011), including data based on the QMAP (Quarter Million Scale deposits found beneath more recent sediments in the European Mapping) project, a nationwide geological mapping project where Alps (e.g. Hinderer, 2001). the ages of quaternary deposits were mapped in terms of marine In summary, recent field based observations and modelling isotope stages (Heron, 2014). As such, we use the mapping of Barrell studies provide evidence that LGM glaciers often overrode pre- (2011) as the input for REVOLTA, with some minor adjustments existing sediments to some extent and did not always reach made to account for new and unexploited evidence, as summarised bedrock. Whilst it may seem surprising that gravels were able to in Table 1. The final extent outline we used can be seen in Fig. 1 and survive the entirety of the LGM, Shulmeister et al. (2019) suggests a full discussion for each individual adjustment can be found in that rapid water drainage through these stratified gravels resulted James (2016). in motion being initially limited to the slow and inefficient process of internal ice deformation, allowing ice advance without sub- 2.3. Centreline generation and point ice thickness estimation stantial erosion of the underlying substrate. As such, whilst techniques exist to estimate and remove the entirety of sediment infill REVOLTA initially estimates ice thickness values at regular in- from a DEM (e.g. Harbor and Wheeler, 1992; Jaboyedoff and Derron, tervals along a network of centrelines in a similar manner to other 2005), this approach is not deemed appropriate for the regional 2D (Benn and Hulton, 2010) and 3D (Pellitero et al., 2016) ap- scale modelling of LGM glaciers in the New Zealand Southern Alps. proaches. Centrelines for all major catchments and tributaries were With some sediment surviving, LGM bed elevations lie some- constructed using a GIS based hydrological routing approach, as where between the contemporary DEM and the level that would used in other glaciological studies (e.g. Schiefer et al., 2008). Cen- exist if all postglacial valley-fill were removed. However, since trelines were manually checked and adjusted to best define the inferred LGM sediment thickness is only available for a few select centre of each glacier branch. locations (e.g. McKinnon et al., 2012), there are not enough data to Ice thickness values were subsequently calculated at 50 m in- generate a regional distribution of the LGM bed surface beneath tervals along each centreline, using a perfect plasticity approach Holocene valley fill sediments. Therefore, with the consensus in the similar to that developed by Benn and Hulton (2010) and used by literature suggesting that at least some sediment was present pre- Pellitero et al. (2016). The original basis for this method was LGM, for the regional scale modelling, REVOLTA was applied introduced by Schilling and Hollin (1981), stating that: W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35 25

Table 1 Summary of revisions made to LGM outline used in this study. Locations shown in Fig. 1. SED ¼ Surface Exposure Date.

Location Description of adjustment (Barrell (2011) as baseline). Basis for adjustment

Hope valley 15 km downstream to ‘Glenhope’ moraines IRSL dating (Rother, 2006) Waimakariri Valley 20 km downstream to ‘Otarama’ moraines Mapping and SEDs (Rother et al., 2015) Rakaia Valley 2 km downstream to ‘Tui Creek’ moraine crest. SEDs (Shulmeister et al., 2010) and mapping (Soons and Gullentops, 1973) Cascade Valley 5 km downstream (to LGM shoreline) Proﬁle of lateral moraine (Sutherland et al., 2007) Upper Clutha Valley (Lake Hawea & Lake 10 km upstream (to Mt. Iron moraine) IRSL dating (Wyshnytzky, 2013) and radiocarbon dating Wanaka) (McKellar, 1960). Milford Sound 3 km upstream (to correspond with submarine terminal Mapping and SEDs (Dykstra, 2012). moraine) Waitutu region Adapted to extent of Ward (1988) Mapping (Ward, 1988)

driving stress in mountain glaciers as friction from the valley walls = D ¼ þ Tb f x provides resistance to flow (Paterson, 1994). This effect is hiþ1 hi (1) H ipg commonly incorporated into perfect plasticity models using a ‘shape factor’ (f ) as shown in Equation (1). REVOLTA dynamically Where hi is the ice surface elevation, H is the ice thickness, Tb is the adjusts f depending on the local valley geometry using a modified basal shear stress (equal to the yield stress), f is a shape factor version of the method described by Benn and Hulton (2010): representing the proportion of the driving stress supported by the bed, p is ice density (917 kg m 3), g is gravitational acceleration A f ¼ (2) (9.81 m s 2) and i is a specified interval of distance along the cen- Hp treline (50 m in this study). Where A is half the glacierised area (of the cross section), p is half the glacierised perimeter and H is the ice thickness at the 2.4. Shear stress centreline. Initially, an approximate ice surface elevation and cross section is generated by using a standard f value of 0.8 (Nye, 1965). Basal shear stress (T ) is a critical parameter of Equation (1) as it b From this initial cross section, a ‘valley geometry based’ f value can determines at which point deformation will occur. T is generally b be calculated using Equation (2) which is used to update the ice between 50 and 150 kPa for mountain valley glaciers (Paterson, surface elevation. In a similar approach to that of Benn and Hulton 1994), although it may vary between individual glaciers due to (2010), an iterative process is applied to determine if the valley various factors (e.g. basal sliding, ice viscosity, subglacial defor- geometry based f value adequately describes the updated cross mation). A constant shear stress value was used along the length of section (defined as within an error of ±0.1), recalculating the ice each centreline as this has been shown to adequately reproduce ice surface elevation and valley geometry based f value until an thickness estimates along the length of a centreline (Li et al., 2012) appropriate solution is found. and has been successfully used in other palaeoglacier reconstruction studies (Rea and Evans, 2007). We calibrated Tb using independent estimates of ice thickness from a variety of published 2.6. Creation of an ice surface records. This method is more reliable than using globally averaged values (e.g. Locke, 1995; Rea and Evans, 2007) and is the preferred For each point at which the ice surface elevation is estimated choice for calibrating perfect plasticity based models (Benn and along the centreline, a ‘nearest neighbour’ approach is used to find Hulton, 2010; Pellitero et al., 2016). the closest point on each of the valley sides which is of equal or Although the standard perfect plasticity model assumes motion greater elevation. The procedure is executed twice; once for each by internal deformation only, basal sliding may also contribute to valley side, and the resultant nearest points effectively mark the motion if glaciers are wet based. It is unclear to what extent basal lateral extent of the glacier being reconstructed. These points on sliding occurred during the LGM in New Zealand, with Golledge the valley sides are assigned a zero ice thickness and are used with et al. (2012) suggesting the geomorphological evidence noted by the centreline point ice thickness values to generate a fully 3D Mager and Fitzsimons (2007) (striated and abraded bedrock, distributed ice surface. deformed subglacial till and the transport of erratics through subglacial pathways) as evidence of its occurrence. Conversely, 2.7. Automated palaeo-ELA estimation Shulmeister et al. (2019) propose that water rapidly drained through the stratified gravel beds, minimizing basal sliding. In The LGM ice surface generated by REVOLTA was used to esti- either case, basal sliding is accounted for in this study as Tb is mate Equilibrium Line Altitudes (ELAs) during the LGM. Firstly, calibrated from local landform evidence, effectively prescribing a individual glacier catchments were approximated by generating softer ice rheology, as described by Veen (2013). watersheds from the newly created LGM ice surface. ELAs for each Whilst prescribing a softer ice-rheology in this manner is suit- individual glacier were then calculated using the Area-Altitude fi able for reconstructing former pro les, it should be noted that Tb Balance Ratio (AABR) method (Osmaston, 2005). This was ach- effectively becomes a ‘lumped parameter’ and further calculations, ieved using Python code modified from Pellitero et al. (2015) and as such as estimates of ice velocity may be spurious. Basal sliding also used by Carrivick et al. (2019) for Little Ice Age glacier ablation needs to be carefully considered if Tb is calculated differently (e.g. areas, but in this case using the newly generated LGM ice surface from globally averaged values). and outlines as inputs. With insufficient data (i.e. mass balance) available to parameterize the balance ratio directly, we used 2.5. Shape factor representative balance ratios suggested by Rea (2009), as recom- mended by Pellitero et al. (2015). To assess the sensitivity of the It is unrealistic to assume that the bed shear stress equals the specific balance ratios used, we calculated the ELAs under two 26 W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35 scenarios; firstly using a ‘global’ value of 1.75 for all glaciers and The results of the shear stress parameterisation are shown for secondly using a ‘mid-latitude maritime’ value (1.9) for glaciers to each location in Fig. 3. A value of 30 kPa best-fitted the available the west of the main divide and a continental mid-latitude value evidence, providing an acceptable fit to all locations. Since these (1.59, derived from the European Alps) for those to the east. The catchments include examples from both the east and west of the values in the latter scenario were chosen to best accommodate the Main Divide, it appears that for the purposes of our study, a single east-west climate gradient of the Southern Alps (Henderson and regionally-averaged value of 30 kPa is appropriate for the entire Thompson, 1999), which is also believed to have been a promi- Southern Alps. Whilst it is acknowledged this approach does not nent during the LGM (Drost et al., 2007; Shulmeister et al., 2019). accommodate some inter-catchment variability, we consider it appropriate for our modelling because our aim is to provide geo- 3. Results metric and volumetric details, rather than information on local ice dynamics or thermal regime, for example. 3.1. Contemporary lake bathymetry 3.3. Distributed ice thickness and total ice volume Table 2 details the lakes removed from the DEM and summary statistics. The location of each lake can be seen on Fig. 1. The lakes Fig. 4a shows the results of the reconstructed LGM ice thickness have a combined volume of 240.6 km3, highlighting the importance across the Southern Alps, with an estimated total volume of of their consideration for palaeo ice thickness estimation. 6,800 km3. The REVOLTA model produces large, predominantly Furthermore, this is the first known attempt to quantify the volume valley-constrained, glaciers flowing to the east of the Main Divide of these lakes, revealing that Lake Wakatipu is volumetrically the (e.g. Rakaia and Pukaki Glaciers, Fig. 4b and c) and some piedmont largest (64.2 km3). This is interesting as the title of the ‘largest lake style glaciers coalescing on the West Coast (e.g. Haast region, of the South Island’ is sometimes noted as Lake Te Anau as it is the Fig. 4d). With the LGM shoreline estimated 125m below present largest by surface area (e.g. McLintock, 1966). (Milne et al., 2005), only 1 glacier extends marginally past the LGM coastline (Fig. 4d), accounting for just 0.5 km3 of ice (<0.007% of the 3.2. Shear stress total modelled LGM ice volume). A large number of nunataks are present, especially around the The vertical extent of LGM glaciation in New Zealand is unde- high peaks of the central region. Far fewer nunataks are present fined for the majority of individual catchments. The available evi- south of the Lake Wakatipu region and indeed REVOLTA produces a dence, where geomorphological mapping (e.g. Barrell et al., 2011) localised ice cap in the Fiordland region (Fig. 4e). A maximum has been combined with geochronological dating and some nu- modelled ice thickness of 1,108 m is located at the Wakatipu Glacier merical modelling, is summarised in Table 3. This dataset permitted (see Fig. 4 for location). multiple model runs of varying shear stress to be made of each LGM glacier to manually best-fit modelled palaeo ice surfaces to that 3.3.1. Sensitivity analysis e shear stress fi eld-measured and dated. Any surface exposure dates cited here- We conducted a sensitivity analysis of Tb using the LGM Pukaki after have been recalibrated to the updated production rate from Glacier as a test case. Analysis was performed for the Pukaki Glacier Putnam et al. (2010) in a similar manner to that of Darvill et al. due to the abundance of dated landform evidence (Table 3). As ‘ fi ’ (2016). previously noted, the best t Tb value of 30 kPa was purposely We aimed to generate a profile slightly above the dated land- chosen to generate a profile slightly above the dated landform ev- form evidence due to the general convex profile of glaciers in the idence. This is due to the general convex profile of glaciers in the ablation zone (Nesje, 1992) where centreline ice elevations are ablation zone (Nesje, 1992) where centreline ice elevations are above those on the valley sides, with a differential of 50 m above those on the valley sides, with a differential of 50 m approximated in previous research deemed appropriate (Glasser approximated in previous research deemed appropriate (Glasser fi ‘ ’ et al., 2011). et al., 2011). Considering such a convex pro le, a low Tb estimate

Table 2 Lakes removed from the contemporary DEM and reference for original source. Summary statistics calculated from interpolated bathymetry.

Lake Name Reference Volume km3 Max depth m Area km2

Te Anau Irwin (1971) 48.25 417 344.2 Hooker Robertson et al. (2012) 0.06 140 1.22 Maud Warren and Kirkbride (1998) 0.08 90 1.40 Godley Warren and Kirkbride (1998) 0.08 90 1.99 Mueller Robertson et al. (2012) 0.02 83 0.87 Pukaki (Irwin, 1970b) 8.9 98 172.8 Rotoroa Irwin (1982b) 2.17 145 23.6 Tasman Dykes et al. (2010) 0.54 240 5.96 Tekapo (Irwin, 1973) 6.94 120 96.5 Monowai Irwin (1981b) 2.56 161 32.0 McKerrow Irwin (1983) 1.69 121 22.9 Manapouri Irwin (1969) 22.47 444 138.6 Kaniere Irwin (1982a) 1.38 198 14.7 Hawea Irwin (1975) 24.49 384 151.7 Hauroko Irwin (1980) 10.97 462 71.0 Ohau (Irwin, 1970a) 4.56 129 59.3 Sumner Irwin (1979) 1.1 134 13.7 Coleridge Flain (1970) 3.65 200 36.9 Brunner Irwin (1981a) 2.24 109 40.6 Wanaka Irwin (1976) 31.92 311 198.9 Wakatipu (Irwin, 1972) 64.2 380 295.4 W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35 27

Table 3 Locations used for manually parameterising shear stress value for use in REVOLTA. See Fig. 1 for locations. SED ¼ Surface Exposure Date.

Location Description Reference(s)

Pukaki Valley Clearly deﬁned 40 km lateral moraine sequence constrained to Doughty et al. (2015) LGM period (18e30 ka) by over 50 SED samples Kelley et al. (2014) Schaefer et al. (2006) Glaciological model McKinnon et al. (2012) Ohau Valley Series of SED samples constraining lateral Putnam et al. (2013) moraines to LGM period (18e30 ka). Glaciological model Putnam et al. (2013) Milford Sound Series of SED samples between 8 and 238 m a.s.l. Dykstra (2012) within LGM period (18e30 ka) Waimakariri Valley SED samples constraining various geological features to Rother et al. (2015) LGM period (18e30 ka) Cascade Valley Clearly deﬁned 13 km long lateral moraine dated to LGM (~21.85 ka) Sutherland et al. (2007) by a series of cosmogenic samples.

(15 kPa) was used to approximate the lowest elevation landform localised ELA low points found where additional mountain masses ‘ ’ evidence whilst a high Tb estimate (40 kPa) was used to exceed the are parallel and offset to the west of the main divide (e.g. Solution elevation of the geomorphological evidence by over 50 m. Fig. 5 Range, Fig. 7a) whilst higher ELAs are found where low passes shows the impact of altering Tb within the bounds described: a Tb penetrate the main divide (e.g. Hollyford Pass, Fig. 7a). Both balance of 15 kPa (Fig. 5a) produces an area 17.5% less than the ‘best fit’ ratio scenarios produce similar results, although a marginally 30 kPa scenario whilst the 40 kPa scenario (Fig. 5c) produces an stronger east-west differential and associated ELA gradient are area 9.5% greater than the baseline. For ice volume there was an evident under the 1.9 west/1.59 east scenario. underestimate of 47.4% (15 kPa model) and an overestimate of LGM ELAs were depressed by approximately 950 m on average, 30.5% (40 kPa model). Much of these differences are driven by the although depressions were greater for glaciers to the west of the specific topography of the Pukaki Valley, although they provide a main divide (under both balance ratio scenarios). Calculation of first order estimate of the model sensitivity. average ELA gradients across three transects (as originally used by Chinn and Whitehouse (1980)) reveals that east-west ELA gradients 3.3.2. Sensitivity analysis e post LGM sediment infill during the LGM were markedly reduced at locations where they are Sensitivity analysis was also conducted with regards to post presently very high (Fig. 7, T2: Mt. Cook and T3: Olivine Range LGM sediment infill. Analysis was performed for the Pukaki Valley transects) whilst increasing at the northernmost transect (T1: as this is the only location of the Southern Alps where sufficient Whitcombe Pass) where a lower ELA gradient is currently present. data on sediment thickness exists (McKinnon et al., 2012). Using the methods described in Section 2.1.5.1, Fig. 6a shows the esti- 4. Discussion mated post-LGM sediment thickness distribution (and areal extent) fi within the Pukaki Valley, revealing up to 348 m of post-LGM in ll. 4.1. Style and nature of glaciation during the LGM Modelled LGM ice thickness is shown using the original (no sedi- ‘ ment removed) DEM in Fig. 6b and for the newly created sediment Our estimate of 6,800 km3 ice volume for the Southern Alps ’ removed version in Fig. 6b. Corresponding cross sections of the during the LGM is similar to that by Golledge et al. (2012) of fi LGM bed pro le, original DEM and ice surfaces are shown in Fig. 6d. 6,400 km3 (adjusted herein for the same spatial extent). The com- For this modelling, all other parameters were set as per the stan- parison is especially interesting because of the markedly different dard model (Tb of 30 kPa). When considering overall ice volume in approaches: Golledge et al. (2012) took climate proxy data to drive the study region, modelling with the standard DEM predicts a fl 3 an ice ow dynamics model, whereas we utilised geomorphological volume 4.0 km (7.3%) lower than that with the sediment removed. evidence of glacier extent to derive distributed glacier ice thickness independent of climate estimates. With current ice volume of 3.4. ELA reconstruction approximately 50 km3 (Chinn, 2001), the Southern Alps has lost approximately 99.25% of its glacial ice since the LGM, highlighting To compare LGM ELAs with contemporary ELAs, we use a digi- the climatic sensitivity of the glaciers. For comparison, Greenland tized version of the contemporary ELA trend surface produced by has lost approximately 36% of its ice since the LGM and Antarctica Chinn and Whitehouse (1980). We use this surface as our baseline just 7% (Bintanja et al., 2002). as it is effectively an ELA0 surface as most glaciers were close to zero There is some debate and uncertainty surrounding the vertical balance during this period (Chinn et al., 2005). Table 4 summarises ice extent in New Zealand during the LGM although there is a the results of the reconstructed LGM ELAs under both balance ratio growing consensus that glaciers were somewhat constrained by scenarios (as described in section 2.7) alongside the contemporary topography and a full ice cap was not present. Anderson and values for corresponding regions, whilst Fig. 7 graphically displays Mackintosh (2006) suggest the abundance of extremely sharp the results for the balance ratio 1.9 west/1.59 east scenario. LGM ridgelines as evidence they were not overridden by ice whilst ELAs were substantially lower than present, with regionally aver- modelling of individual catchments clearly shows exposed ridge- aged estimates of approximately 900 m. There was large spatial lines around the Mt. Cook region (Putnam et al., 2013). Although variability of ELAs during the LGM, as there is at present (Carrivick the previous regional modelling by Golledge et al. (2012) presents a and Chase, 2011), with LGM glaciers flowing to the west coast continuous ‘icefield’ style of glaciation, ice thickness over many of exhibiting ELAs approximately 320 m lower than those to the east. the peaks and ridgelines is actually modelled to be extremely low. A gradual south west e north east trend can also be observed The results of REVOLTA in this study are consistent with a valley (Fig. 7a), with the lowest values found in the south west. Super- constrained style of glaciation, with exposed ridges and nunataks, imposed upon these trends is further local variability, with especially in central and northern regions (Fig. 4). The main 28 W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35

Fig. 3. Optimal surface proﬁles (30 kPa) and independent evidence of vertical extent. Blue triangles/lines show the elevation of landforms dated to the LGM and are therefore suitable ‘targets’ for reconstruction. SED ¼ Surface Exposure Dates. (For interpretation of the references to colour in this ﬁgure legend, the reader is referred to the Web version of this article.) W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35 29

Fig. 4. a) REVOLTA derived 100 m resolution simulation of LGM ice thickness for the Southern Alps domain using optimal parameters (Tb ¼ 30 kPa). 3D simulations of: b) Rakaia Glacier, c) Pukaki/Ohau glaciers, d) Haast Glacier/west coast and e) Fiordland ‘Iceﬁeld’. 30 W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35

Fig. 5. LGM Pukaki Glacier modelled under different scenarios: a) 15 kPa, b) 30 kPa, c) 40 kPa. exception to this is in the Fiordland region, where REVOLTA sug- presented in Fig. 3 are similar to those of Golledge et al. (2012), the gests a localised ice cap style of glaciation (Fig. 4e). Interestingly, Milford profile (Fig. 3e) is markedly different, with differentials of this pattern is almost identical to that proposed by Chinn et al. ice surface elevation in excess of 300 m in places. The good agree- (2014) who postulated that the Darran Range (adjacent to Milford ment between REVOLTA and the (albeit limited) surface exposure Sound) marked an approximate boundary between a southern ice dates in this catchment suggests that the results of REVOLTA are the cap and a northern region where peaks and ridges penetrated the most plausible in this location. One potential reason for Golledge LGM ice. et al. (2012) overestimating glacier ice surface elevations and vol- The REVOLTA modelling approach also provides an indirect ume in this specific catchment (compared the age controlled insight into potential basal conditions as the parameterisation geomorphological evidence and REVOLTA) is their assumption of scheme for Tb effectively accounts for basal sliding (Veen, 2013). spatially-uniform differentials of temperature and precipitation fi Speci cally, we found a Tb value of 30 kPa best represented land- between the LGM and present. This is noted in their analysis, where form evidence, with a range between 15 kPa and 40 kPa deemed scenarios which better reproduced known glacial limits to the east plausible. If motion was solely by internal deformation we would of the main divide (e.g. the Rakaia) produced overextended glaciers e ‘ expect Tb to be around 50 150 kPa (Nye, 1952; Paterson, 1994). As in the central west coast areas. In essence, their optimal recon- such, we interpret that basal sliding was prominent for NZ glaciers struction’ can be seen as a compromise between under-estimating during the LGM. Whilst this is in agreement with other studies (e.g. certain glaciers to the east of the main divide (as discussed in their Golledge et al., 2012) due to sedimentary and geomorphological analysis and shown in Fig. 1) and over-estimating those of the evidence (Mager and Fitzsimons, 2007; Evans et al., 2013), central west coast (as shown in Fig. 3e). Shulmeister et al. (2019) suggests this is not always the case. In their study, it is suggested that basal substrate plays a crucial role, 4.2. ELA reconstruction and implications for climatic conditions with rapid water drainage through stratified gravels limiting motion to internal deformation. Conversely, when the bed consists of a Our LGM ELA estimates for individual glaciers of between 375 m high mud content (e.g. lake basins), rapid basal sliding may occur. and 1,502 m are of a similar range to those reported by Golledge Whilst our study supports the notion of generally wet and sliding et al. (2012) (approximately 350 me1,450 m) and show an overall conditions at the regional scale, further research is needed to similar pattern. The two balance ratio scenarios we used produced quantify variability over space and time. an almost identical regionally averaged ELA (900 m compared to Whilst the total volume and most of the longitudinal profiles 898 m) whilst the 1.9 west/1.59 east scenario produced a slightly W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35 31

Fig. 6. a) Estimated areal extent and thickness of post LGM-sediment in the Pukaki basin. b) Modelled ice thickness (standard DEM). c) Modelled ice thickness (sediment removed DEM). d) Cross-section of M1-M2 transect.

Table 4 Contemporary and LGM ELAs for the Southern Alps, associated depression during the LGM and east-west gradients. See Fig. 7a for location of transects.

Contemporary LGM

balance ratio 1.75 balance ratio 1.9 west/1.59 east

Mean (entire) 1,849 m 900 m 898 m Mean (east) 1,976 m 1,087 m 1,100 m Mean (west) 1,681 m 774 m 760 m Depression from contemporary (entire) 949 m 951 m Depression from contemporary (east) 889 m 876 m Depression from contemporary (west) 907 m 921 m ELA gradient, transect T1 (Whitcombe Pass) 3.7 m/km 1 11.6 m/km 1 12.6 m/km 1 ELA gradient, transect T2 (Mt. Cook Region) 31 m/km1 13.5 m/km1 14.0 m/km1 ELA gradient, transect T3 (Olivine Range) 20 m/km 1 12.1 m/km 1 12.5 m/km 1

increased east/west differential, as would be expected (Table 4). As balance ratios on each side of the main divide. the differences are minimal, we suggest that either scenario is A high east-west ELA gradient has long been recognized be- appropriate for the purposes of our modelling, although the 1.9 tween contemporary Southern Alps glaciers and is often inter- west/1.59 east scenario may be preferred as this better represents preted as a response to the current precipitation pattern (Porter, the known conditions of the glaciers currently in the region (Chinn 1975a). As such we infer the presence of an east-west ELA et al., 2005). Importantly, the global scenario (balance ratio for all gradient during the LGM as evidence that westerly circulation was glaciers ¼ 1.75) still reproduces a strong west-east differential, so in operation to some extent during the LGM, a view shared by other we can be conﬁdent that this differential is not due to the choice of authors (e.g. Rother and Shulmeister, 2006). Crucially however, our 32 W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35

Fig. 7. a) LGM ELAs (AABR, balance ratio 1.9 west/1.59 east). b) LGM ELA depression from present. results show that there was spatial variability in the ELA depression contemporary distribution (Landcare Research Ltd, 2003a). By dif- and associated gradients, suggesting that climatic change was not ferencing the LGM estimates and corresponding contemporary uniform across the Southern Alps. Analysis of the ‘Mt. Cook’ and datasets, the differentials can be calculated for both temperature ‘Olivine Range’ transects (Fig. 7) shows that LGM ELA gradients (Fig. 8e) and precipitation (Fig. 8f). Note that LGM glaciers are were markedly reduced compared to present (reducing from 31 to delineated in Fig. 8e as no correction has been made for changes in 14 m/km 1 and 20 and 12.5 m/km 1 respectively) whilst the ice surface elevation and associated effect of lapse rate. Importantly, currently low ELA gradient at the Whitcombe Pass (3.7 m/km 1) the downscaling precipitation algorithm used by Karger et al. was enhanced during the LGM to 12.6 m/km 1. (2017) and Brown et al. (2018) incorporates orographic predictors Our assertion of spatial variability in climatic change is in line including wind fields, valley exposition, and boundary layer height, with a growing body of evidence from other studies using a variety with a subsequent bias correction. We are therefore confident that of techniques, although to our knowledge this study is the first to the modelled precipitation data (Fig. 8c) adequately captures the use glacial reconstruction for such a purpose in New Zealand. magnitude of orographic precipitation enhancement in the high- McKinnon et al. (2012) collated estimates of temperature deviation relief setting of the Southern Alps. (for the New Zealand LGM) from different studies, finding a large These new, high resolution datasets form an independent range in reported values. For example, Sikes et al. (2002) found that source for assessing the spatial variability of climatic parameters. sea surface temperature south of Chatham rise were 8 C cooler Fig. 8e and f demonstrates that changes in temperature and pre- than present whilst Marra et al. (2006) inferred very moderate cipitation distribution varied spatially, as predicted by the ELA re- cooling in the Canterbury region, as little as 2 C on average (albeit constructions of this study (Fig. 7). Fig. 8e shows the majority of the probably during a short inter-stadial within the extended LGM). Southern Alps experienced cooling in the range of 2 e 5 C, broadly The notion of spatial variability is supported by climate modelling similar to the New Zealand average of 4.6 C reported by Drost et al. by Drost et al. (2007) who estimated that temperature reduction (2007) although perhaps slightly lower than the 6.5 C suggested varied spatially between 2.5 C and 4 C from pre-industrial levels. by Golledge et al. (2012). Whilst the negligible cooling estimated in Although LGM precipitation estimates are less well constrained north-west regions is surprising, this is within the ranges reported than temperature, there is evidence to suggest that both the in the literature, with Marra et al. (2006) suggesting a temperature amount and distribution were significantly different from present. increase of up to 0.5 C during summer LGM interstadials at Lyndon Palaeo-vegetation studies suggest a major reorganization of vege- Stream (see Fig. 8e for location) whilst Samson et al. (2005) re- tation patterns during the New Zealand LGM, with a general ported winter sea surface temperature decreases of just 0.9 C consensus of drier conditions overall (McGlone et al., 1993). Pre- (offshore of New Zealand North Island). cipitation patterns were also likely to be different during the LGM, For precipitation, Fig. 8f shows major changes in the amount and with recent research by Shulmeister et al. (2019) postulating distribution during the LGM. Whilst an east-west precipitation changes driven by a shift in the track of westerlies. Precipitation gradient is clearly still modelled during the LGM (Fig. 8c), this is modelling by Drost et al. (2007) also supports the concept of a much reduced, with less precipitation on the west coast, which generally drier climate, albeit with substantial local variation. currently has some of the highest rainfall on Earth (Henderson and Most recently, downscaled climate modelling by Karger et al. Thompson, 1999). The pattern is consistent with the assertion of an (2017) and Brown et al. (2018) has produced high resolution overall drier conditions during the LGM (McGlone et al., 1993; Drost (~1 km) gridded estimates of LGM mean annual temperature and et al., 2007). Drost et al. (2007) also found the greatest precipitation precipitation, which were downloaded from www.paleoclim.org. decreases in Western regions, with average reductions of 30% (from Fig. 8a displays the mean annual temperature distribution at the pre-industrial levels) in Southern Otago-Southland regions. LGM using the CCSM4 model, whilst Fig. 8b shows the corre- In essence, the climate model estimates based on the work of sponding values at present (Landcare Research Ltd, 2003b). Similar Karger et al. (2017) and Brown et al. (2018) are in correspondence maps are shown for precipitation distribution, with Fig. 8c showing with the results from the ELA reconstructions in this study (Fig. 7a): the modelled LGM distribution and Fig. 8d showing the namely a weakened (albeit still important) east-west climatic W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35 33

Fig. 8. Changes in climate (mean annual temperature and precipitation) during the LGM based on modelled LGM climate (CCSM4 model, downloaded from www.paleoclim.org) and contemporary values (Landcare Research Ltd, 2003a, 2003b). gradient with substantial local level variability. The agreement is divide. especially interesting because of the markedly different ap- The recognition of spatially variable climatic differentials is proaches: Karger et al. (2017) and Brown et al. (2018) used a global important because they have not been previously considered when climate model to estimate temperature and precipitation wheras reconstructing LGM ice thickness in New Zealand. Namely, the only we utilised geomorphological evidence to derive ice thickness and previous such model by Golledge et al. (2012) assumed spatially- associated ELAs. uniform differentials of temperature and precipitation between the LGM and present. Whilst producing an acceptable fit at the regional scale, they were unable to resolve known ice limits in some 5. Conclusions catchments (by up to 40 km). We suggest that some of these anomalies may be due to the use of unrealistic uniform climate In this study we used geomorphological and geochronological differentials, especially precipitation distribution to which glaciers records from the Southern Alps of New Zealand to drive a glacial of the Southern Alps are known to be highly sensitive (Rowan et al., reconstruction model (REVOLTA) for the LGM. We modelled a total 2014). The advent of independent high resolution estimates of volume of 6,800 km3 of glacier ice across the Southern Alps during palaeo climatic conditions (Karger et al., 2017; Brown et al., 2018) the LGM, representing a 99.25% volume loss to present. Analysis of may help to address some of these issues, allowing further differ- the distributed ice thickness reveals a ‘valley-constrained' style of ences between ‘top down’ climate based reconstructions and glaciation with many exposed ridges and nunataks, even when ‘bottom up’ landform based reconstructions (e.g. REVOLTA) to be parameterised to fit the maximum vertical extent of geomorpho- assessed in more detail. The framework presented here could easily logical evidence. An ice-cap style of glaciation was predicted for the be applied to other formerly glaciated regions, helping to improve Fiordland region, where maximum surface elevations are lower. We our knowledge of the global climate system. found a low shear stress value of approximately 30 kPa was required to best fit the geomorphological evidence, from which we infer that at least some glaciers were warm-based during the LGM. Acknowledgements Reconstructed ELAs reveal a regionally averaged depression of approximately 950 m during the LGM. This ELA depression is WJ was funded by a NERC PhD studentship (NE/K500847/1). JC consistent with findings of other studies. However, and crucially, and WJ thanks the School of Geography, University of Leeds, for fi we found substantial spatial variability in ELA depression, with assistance with eldwork costs. Trevor Chinn commented on an east-west ELA gradients during the LGM generally decreased. In early version of the manuscript. conjunction with evidence from the latest independent climate models (Karger et al., 2017; Brown et al., 2018), we propose that Appendix A. Supplementary data westerly circulation was decreased during the LGM compared to at present, resulting in decreased precipitation to the west of the main Supplementary data to this article can be found online at 34 W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35 https://doi.org/10.1016/j.quascirev.2019.06.035. Institute Chart Series, p. 760. Geographix, 2012. NZ 8m Digital Elevation Model. www.geographx.co.nz. (Accessed January 2018). References Glasser, N.F., Harrison, S., Jansson, K.N., Anderson, K., Cowley, A., 2011. Global sea- level contribution from the patagonian icefields since the little ice age Adams, J., 1980. Contemporary uplift and erosion of the southern Alps, New Zea- maximum. Nat. Geosci. 4, 303e307. land’. Geol. Soc. Am. Bull. 91, 1e114. Golledge, N.R., Mackintosh, A.N., Anderson, B.M., Buckley, K.M., Doughty, A.M., Alexander, D., Davies, T., Shulmeister, J., 2014. Formation of the Waiho Loop ter- Barrell, D.J.A., Denton, G.H., Vandergoes, M.J., Andersen, B.G., Schaefer, J.M., minal moraine, New Zealand. J. Quat. Sci. 29, 361e369. 2012. Last glacial maximum climate in New Zealand inferred from a modelled Anderson, B.M., Hindmarsh, R.C.A., Lawson, W.J., 2004. A Modelling Study of the southern Alps icefield. Quat. Sci. Rev. 46, 30e45. Response of Hatherton Glacier to Ross Ice Sheet Grounding Line Retreat’, Global Harbor, J.M., Wheeler, D.A., 1992. On the mathematical description of glaciated and Planetary Change, vol. 42. Elsevier, pp. 143e153, 1e4. valley cross sections. Earth Surf. Process. Landforms 17, 477e485. Anderson, B., Mackintosh, A., 2006. Temperature change is the major driver of late- Henderson, R.D., Thompson, S.M., 1999. Extreme rainfalls in the southern Alps of glacial and Holocene glacier fluctuations in New Zealand. Geology 34, 121e124. New Zealand. J. Hydrol. (New Zealand) 38, 309e330. Barrell, D.J.A., Andersen, B.G., Denton, G.H., Smith Lyttle, B., 2011. Glacial Geo- Heron, D.W., 2014. ‘Geological Map of New Zealand 1:250 000’, GNS Science morphology of the Central South Island, New Zealand, vol. 27. GNS Science Geological Map 1. GNS Science, New Zealand. Lower Hutt. monograph. Hinderer, M., 2001. Late Quaternary denudation of the Alps, valley and lake fillings Barrell, D.J.A., 2011. Quaternary glaciers of New Zealand. In: Ehlers, J., Gibbard, P., and modern river loads. Geodin. Acta 14, 231e263. Hughes, P. (Eds.), Developments in Quaternary Science. Elsevier, Amsterdam. Hutchinson, M.F., 1989. A new procedure for gridding elevation and stream line Beavan, J., Ellis, S., Wallace, L., Denys, P., 2007. ‘Kinematic Constraints from GPS on data with automatic removal of spurious pits. J. Hydrol. 106, 211e232. Oblique Convergence of the Pacific and Australian Plates, Central South Island, Irwin, J., 1969. Lake Manapouri Bathymetry, 1:31,680. New Zealand Oceanographic New Zealand’, vol. 175. Geophysical Monograph Series, pp. 75e94. Institute Chart Series. Benn, D.I., Hulton, N.R.J., 2010. An Excel (TM) spreadsheet program for recon- Irwin, J., 1970a. Lake Ohau Bathymetry, 1:31,680. New Zealand Oceanographic structing the surface profile of former mountain glaciers and ice caps. Comput. Institute Chart Series. Geosci. 36, 605e610. Irwin, J., 1970b. Lake Pukaki bathymetry, 1:31,680. New Zealand Oceanographic Bintanja, R., Van de Wal, R.S.W., Oerlemans, J., 2002. Global ice volume variations Institute Chart Series. through the last glacial cycle simulated by a 3-D ice-dynamical model. Quat. Int. Irwin, J., 1971. Lake Te Anau Bathymetry, 1:63,360. New Zealand Oceanographic 95, 11e23. Institute Chart Series. Brown, J.L., Hill, D.J., Dolan, A.M., Carnaval, A.C., Haywood, A.M., 2018. PaleoClim, Irwin, J., 1972. Lake Wakatipu Bathymetry, 1:63,360. New Zealand Oceanographic High Spatial Resolution Paleoclimate Surfaces for Global Land Areas, vol. 5. Institute Chart Series. Scientific data, 180254. Irwin, J., 1973. Lake Tekapo Bathymetry, 1:31,680. New Zealand Oceanographic Burdis, A.J., 2014. Denudation Rates Derived from Spatially-Averaged Cosmogenic Institute Chart Series. Nuclide Analysis in Nelson/Tasman Catchments, South Island, New Zealand. Irwin, J., 1975. Lake Hawea Bathymetry, 1:47,520. New Zealand Oceanographic PhD thesis. University of Wellington. Institute Chart Series. Carrivick, J.L., Davies, B.J., James, W.H.M., Quincey, D.J., Glasser, N.F., 2016. Distrib- Irwin, J., 1976. Lake Wanaka Bathymetry, 1:55,440. New Zealand Oceanographic uted ice thickness and glacier volume in southern South America. Global and Institute Chart Series. Planetary Change 146, 122e132. Irwin, J., 1979. Lake Sumner, Katrine, Mason and Marion Bathymetry, 1:12,000. New Carrivick, J.L., Davies, B.J., James, W.H.M., McMillan, M., Glasser, N.F., 2019. Zealand Oceanographic Institute Chart Series. A comparison of modelled ice thickness and volume across the entire Antarctic Irwin, J., 1980. Lake Hauroko Bathymetry, 1:30,000. New Zealand Oceanographic Peninsula region, Geografiska Annaler: Series A. Physical Geography 101 (1), Institute Chart Series. 45e67. Irwin, J., 1981a. Lake Brunner Bathymetry, 1:15,000. New Zealand Oceanographic Carrivick, J.L., Boston, C.M., King, O., James, W.H.M., Quincey, D.J., Smith, M.W., Institute Chart Series. Grimes, M., Evans, J., 2019. Accelerated volume loss in glacier ablation zones of Irwin, J., 1981b. ‘Lake Monowai Bathymetry, 1:20,877’. New Zealand Oceanographic NE Greenland, Little Ice Age to present. In: Geophysical Research Letters, vol. 46. Institute Chart Series. American Geophysical Union. https://doi.org/10.1029/2018GL081383. Irwin, J., 1982a. Lake Kaniere Bathymetry, 1:12,000. New Zealand Oceanographic Carrivick, J.L., Chase, S.E., 2011. Spatial and temporal variability in the net mass Institute Chart Series. balance of glaciers in the Southern Alps, New Zealand. N. Z. J. Geol. Geophys. 54, Irwin, J., 1982b. Lake Rotoroa Bathymetry, 1:15,840’. New Zealand Oceanographic 415e429. Institute Chart Series. Chinn, T., Kargel, J.S., Leonard, G.J., Haritashya, U.K., Pleasants, M., 2014. New Zea- Irwin, J., 1983. Lake McKerrow and Lake Alabaster Bathymetry, 1:18,000. New land's glaciers. In: JS, K., et al. (Eds.), Global Land Ice Measurements from Space. Zealand Oceanographic Institute Chart Series. Praxxis, Chichester. Jaboyedoff, M., Derron, M.-H., 2005. A new method to estimate the infilling of al- Chinn, T.J., 2001. Distribution of the glacial water resources of New Zealand. luvial sediment of glacial valleys using a sloping local base level. Geografia J. Hydrol. (New Zealand) 40, 139e187. Fisica e Dinamica Quaternaria 28, 37e46. Chinn, T.J.H., Whitehouse, I.E., 1980. ‘Glacier Snow Line Variations in the Southern James, W.H., Carrivick, J.L., 2016. Automated modelling of distributed glacier ice Alps, New Zealand’, World Glacier Inventory, vol. 126. International Association thickness and volume. Comput. Geosci. 92, 90e103. of Hydrological Sciences Publication, pp. 219e228. James, W.H.M., 2016. A Landform Based 3D Reconstruction of Glacier Ice at the Last Chinn, T.J., Heydenrych, C., Salinger, M.J., 2005. ‘Use of the ELA as a practical method Glacial Maximum the Southern Alps, New Zealand. PhD Thesis. University of of monitoring glacier response to climate in New Zealand's Southern Alps'. Leeds. J. Glaciol. 51, 85e95. Kamp, P.J.J., Green, P.F., White, S.H., 1989. Fission track analysis reveals character of Coates, G., 2002. The Rise and Fall of the Southern Alps. Canterbury University collisional tectonics in New Zealand. Tectonics 8, 169e195. € Press, Christchurch. Karger, D.N., Conrad, O., Bohner, J., Kawohl, T., Kreft, H., Soria-Auza, R.W., Darvill, C.M., Bentley, M.J., Stokes, C.R., Shulmeister, J., 2016. The timing and cause of Zimmermann, N.E., Linder, H.P., Kessler, M., 2017. Climatologies at High Reso- glacial advances in the southern mid-latitudes during the last glacial cycle lution for the Earth's Land Surface Areas, vol. 4. Scientific data. Nature Pub- based on a synthesis of exposure ages from Patagonia and New Zealand. Quat. lishing Group (170122). Sci. Rev. 149, 200e214. Kelley, S.E., Kaplan, M.R., Schaefer, J.M., Andersen, B.G., Barrell, D.J.A., Putnam, A.E., Doughty, A.M., Schaefer, J.M., Putnam, A.E., Denton, G.H., Kaplan, M.R., Barrell, D.J.A., Denton, G.H., Schwartz, R., Finkel, R.C., Doughty, A.M., 2014. High-precision Andersen, B.G., Kelley, S.E., Finkel, R.C., Schwartz, R., 2015. Mismatch of glacier 10Be chronology of moraines in the Southern Alps indicates synchronous extent and summer insolation in Southern Hemisphere mid-latitudes. Geology cooling in Antarctica and New Zealand 42,000 years ago. Earth Planet. Sci. Lett. 43, 407e410. 405, 194e206. Drost, F., Renwick, J., Bhaskaran, B., Oliver, H., McGregor, J., 2007. ‘A simulation of Landcare Research Ltd, 2003a. Mean Annual Precipitation Dataset. www.lris.scinfo. New Zealand's climate during the Last Glacial Maximum’. Quat. Sci. Rev. 26, org.nz. (Accessed March 2018). 2505e2525. Landcare Research Ltd, 2003b. Mean Annual Temperature Dataset. www.lris.scinfo. Dykes, R.C., Brook, M.S., Winkler, S., 2010. The contemporary retreat of Tasman org.nz. (Accessed August 2018). Glacier, Southern Alps, New Zealand, and the evolution of Tasman proglacial Levy, R.H., Dunbar, G., Vandergoes, M., Howarth, J.D., Kingan, T., Pyne, A.R., lake since AD 2000. Erdkunde 64, 141e154. Brotherston, G., Clarke, M., Dagg, B., Hill, M., 2018. A High-Resolution Climate Dykstra, J., 2012. The Role of Mass Wasting and Ice Retreat in the Post-LGM Evo- Record Spanning the Past 17 000 Years Recovered from Lake Ohau, South Island, lution of Milford Sound, Fiordland, New Zealand. PhD thesis. University of New Zealand. Scientific Drilling. Canterbury, New Zealand. Li, H., Ng, F., Li, Z., Qin, D., Cheng, G., 2012. An extended “perfect-plasticity” method Evans, D.J.A., Rother, H., Hyatt, O.M., Shulmeister, J., 2013. The glacial sedimentology for estimating ice thickness along the flow line of mountain glaciers. J. Geophys. and geomorphological evolution of an outwash head/moraine-dammed lake, Res. 117 (F1) https://doi.org/10.1029/2011jf002104. South Island, New Zealand. Sediment. Geol. 284, 45e75. Linsbauer, A., Paul, F., Haeberli, W., 2012. Modeling glacier thickness distribution Finlayson, A., 2013. Digital surface models are not always representative of former and bed topography over entire mountain ranges with GlabTop: application of a glacier beds: palaeoglaciological and geomorphological implications. Geo- fast and robust approach. J. Geophys. Res. 117 (F3) https://doi.org/10.1029/ morphology 194, 25e33. 2011JF002313. Flain, M., 1970. Lake Coleridge Bathymetry, 1:23. New Zealand Oceanographic Locke, W.W., 1995. Modelling of icecap glaciation of the northern Rocky Mountains W.H.M. James et al. / Quaternary Science Reviews 219 (2019) 20e35 35

of Montana. Geomorphology 14, 123e130. Rother, H., Shulmeister, J., Fink, D., Alexander, D., Bell, D., 2015. Surface exposure Mager, S., Fitzsimons, S., 2007. Formation of glaciolacustrine late Pleistocene end chronology of the Waimakariri glacial sequence in the Southern Alps of New moraines in the Tasman valley, New Zealand. Quat. Sci. Rev. 26, 743e758. Zealand: implications for MIS-2 ice extent and LGM glacial mass balance. Earth Marra, M.J., Shulmeister, J., Smith, E.G.C., 2006. Reconstructing temperature during Planet. Sci. Lett. 429, 69e81. the last glacial maximum from Lyndon stream, South Island, New Zealand using Rother, H., Shulmeister, J., 2006. Synoptic climate change as a driver of late Qua- beetle fossils and maximum likelihood envelopes. Quat. Sci. Rev. 25, 1841e1849. ternary glaciations in the mid-latitudes of the Southern Hemisphere. Clim. Past Mathews, W.H., 1967. Profiles of late Pleistocene glaciers in New Zealand. N. Z. J. 2, 11e19. Geol. Geophys. 10, 146e163. Rother, H., Shulmeister, J., Rieser, U., 2010. Stratigraphy, optical dating chronology McGlone, M.S., Salinger, M.J., Moar, N.T., 1993. Paleovegetation studies of New (IRSL) and depositional model of pre-LGM glacial deposits in the Hope Valley, Zealand's climate since the last glacial maximum. In: Wright, H.E., et al. (Eds.), New Zealand. Quat. Sci. Rev. 29, 576e592. Global Climates since the LGM. University of Minnesota, pp. 294e316. Rowan, A.V., Brocklehurst, S.H., Schultz, D.M., Plummer, M.A., Anderson, L.S., McKellar, I.C., 1960. Pleistocene deposits of the upper Clutha valley, Otago, New Glasser, N.F., 2014. Late Quaternary glacier sensitivity to temperature and pre- Zealand. N. Z. J. Geol. Geophys. 3, 432e460. cipitation distribution in the Southern Alps of New Zealand. J. Geophys. Res.: McKinnon, K.A., Mackintosh, A.N., Anderson, B.M., Barrell, D.J.A., 2012. The influ- Earth Surface 119, 1064e1081. ence of sub-glacial bed evolution on ice extent: a model-based evaluation of the Samson, C.R., Sikes, E.L., Howard, W.R., 2005. Deglacial paleoceanographic history of Last Glacial Maximum Pukaki glacier, New Zealand. Quat. Sci. Rev. 57, 46e57. the Bay of plenty, New Zealand. Paleoceanography 20 (4). McLintock, A., 1966. An Encyclopaedia of New Zealand. R. E. Owen, Govt. printer, Schaefer, J.M., Denton, G.H., Barrell, D.J.A., Ivy-Ochs, S., Kubik, P.W., Andersen, B.G., Wellington. Phillips, F.M., Lowell, T.V., Schlüchter, C., 2006. Near-synchronous interhemi- Milne, G.A., Long, A.J., Bassett, S.E., 2005. Modelling Holocene relative sea-level spheric termination of the last glacial maximum in mid-latitudes. Science 312, observations from the Caribbean and South America. Quat. Sci. Rev. 24, 1510e1513. 1183e1202. Schiefer, E., Menounos, B., Wheate, R., 2008. An inventory and morphometric Mitchell, J.S., Mackay, K.A., Neil, H.L., Mackay, E.J., Pallentin, A., P, N., 2012. Undersea analysis of British Columbia glaciers, Canada. J. Glaciol. 54, 551e560. New Zealand, 1:5,000,000. www.niwa.co.nz. (Accessed November 2018). Schilling, D.H., Hollin, J.T., 1981. Numerical reconstructions of valley glaciers and Nesje, A., 1992. Topographical effects on the equilibrium-line altitude on glaciers. small ice caps. In: Hughes, P., GH, D. (Eds.), The Last Great Ice Sheets. John Wiley Geojournal 27, 383e391. and Sons, New York, pp. 207e221. Norris, R.J., Cooper, A.F., 2001. Late Quaternary slip rates and slip partitioning on the Shulmeister, J., Thackray, G.D., Rieser, U., Hyatt, O.M., Rother, H., Smart, C.C., Alpine Fault, New Zealand. J. Struct. Geol. 23, 507e520. Evans, D.J.A., 2010. The stratigraphy, timing and climatic implications of gla- Nye, J.F., 1952. A method of calculating the thicknesses of the ice-sheets. Nature 169, ciolacustrine deposits in the middle Rakaia Valley, South Island, New Zealand. 529e530. https://doi.org/10.1038/169529a0. Quat. Sci. Rev. 29 (2), 2362e2381. Nye, J.F., 1965. The flow of a glacier in a channel of rectangular, elliptic or parabolic Shulmeister, J., Thackray, G.D., Rittenour, T.M., Fink, D., Patton, N.R., 2019. The cross-section. J. Glaciol. 5, 661e690. Timing and Nature of the Last Glacial Cycle in New Zealand’, Quaternary Science Osmaston, H., 2005. Estimates of Glacier Equilibrium Line Altitudes by the Area Reviews, vol. 206. Elsevier, pp. 1e20. Altitude, the Area Altitude Balance Ratio and the Area Altitude Balance In- Sikes, E.L., Howard, W.R., Neil, H.L., Volkman, J.K., 2002. ‘Glacial-interglacial sea dex Methods and Their Validation’, Quaternary International, vol. 138. Elsevier, surface temperature changes across the subtropical front east of New Zealand pp. 22e31. based on alkenone unsaturation ratios and foraminiferal assemblages’. Paleo- Paterson, W.S.B., 1994. The Physics of Glaciers, third ed. Pergamon, Oxford. ceanography 17 (2). Pellitero, R., Rea, B.R., Spagnolo, M., Bakke, J., Ivy-Ochs, S., Hughes, P., Lukas, S., Soons, J.M., Gullentops, F.W., 1973. Glacial advances in the Rakaia valley, New Ribolini, A., 2015. A GIS tool for automatic calculation of glacier equilibrium-line Zealand. N. Z. J. Geol. Geophys. 16, 348e425. altitudes. Comput. Geosci. 82, 55e62. Straumann, R.K., Korup, O., 2009. Quantifying postglacial sediment storage at the Pellitero, R., Rea, B.R., Spagnolo, M., Bakke, J., Ivy-Ochs, S., Frew, C.R., Hughes, P., mountain-belt scale. Geology 37, 1079e1082. Ribolini, A., Lukas, S., Renssen, H., 2016. GlaRe, a GIS Tool to Reconstruct the 3D Suggate, R.P., 1965. Late Pleistocene Geology of the Northern Part of the South Is- Surface of Palaeoglaciers', Computers & Geosciences, vol. 94. Elsevier, land, New Zealand. New Zealand Dept. of Scientific and Industrial Research. pp. 77e85. Sutherland, J.L., Carrivick, J.L., Shulmeister, J., Quincey, D.J., James, W.H.M., 2019. Ice- Pickrill, R.A., Fenner, J.M., McGlone, M.S., 1992. Late quaternary evolution of a fjord contact proglacial lakes associated with the last glacial maximum across the environment in preservation inlet, New Zealand. Quat. Res. 38, 331e346. southern Alps, New Zealand. Quat. Sci. Rev. 213, 67e92. Porter, S.C., 1975a. Equilibrium-line altitudes of late quaternary glaciers in the Sutherland, R., Kim, K., Zondervan, A., McSaveney, M., 2007. Orbital forcing of mid- southern Alps, New Zealand. Quat. Res. 5, 27e47. latitude Southern Hemisphere glaciation since 100 ka inferred from cosmo- Porter, S.C., 1975b. ‘Glaciation Limit in New Zealand's Southern Alps', pp. 33e37. genic nuclide ages of moraine boulders from the Cascade Plateau, southwest Arct. Alp. Res. New Zealand. Geol. Soc. Am. Bull. 119, 443e451. Putnam, A.E., Schaefer, J.M., Barrell, D.J.A., Vandergoes, M., Denton, G.H., Thomas, A.M., 2018. Glacio-lacustrine Sedimentation in Newly Discovered Paleo- Kaplan, M.R., Finkel, R.C., Schwartz, R., Goehring, B.M., Kelley, S.E., 2010. In situ Lakes, Westland. Victoria University of Wellington, New Zealand’. cosmogenic 10Be production-rate calibration from the Southern Alps, New Tippett, J.M., Kamp, P.J.J., 1995. Quantitative relationships between uplift and relief Zealand. Quat. Geochronol. 5, 392e409. parameters for the Southern Alps, New Zealand, as determined by fission track Putnam, A.E., Schaefer, J.M., Denton, G.H., Barrell, D.J.A., Birkel, S.D., Andersen, B.G., analysis. Earth Surf. Process. Landforms 20, 153e175. Kaplan, M.R., Finkel, R.C., Schwartz, R., Doughty, A.M., 2013. The last glacial Trommelen, M., Ross, M., 2010. Subglacial landforms in northern Manitoba, Canada, maximum at 44 S documented by a 10Be moraine chronology at Lake Ohau, based on remote sensing data. J. Maps 6, 618e638. southern Alps of New Zealand. Quat. Sci. Rev. 62, 114e141. Veen, C.J., 2013. Fundamentals of Glacier Dynamics, second ed. Balkema, Rotterdam. Rea, B.R., 2009. Defining modern day Area-Altitude Balance Ratios (AABRs) and Ward, C.M., 1988. Marine terraces of the Waitutu district and their relation to the their use in glacier-climate reconstructions. Quat. Sci. Rev. 28 (3e4), 237e248. late Cenozoic tectonics of the southern Fiordland region, New Zealand. J. R. Soc. Rea, B.R., Evans, D.J.A., 2007. Quantifying climate and glacier mass balance in north N. Z. 18, 1e28. Norway during the Younger Dryas. Palaeogeogr. Palaeoclimatol. Palaeoecol. 246, Warren, C.R., Kirkbride, M.P., 1998. ‘Temperature and bathymetry of ice-contact 307e330. lakes in mount Cook national park, New Zealand’. N. Z. J. Geol. Geophys. 41, Robertson, C.M., Benn, D.I., Brook, M.S., Fuller, I.C., Holt, K.A., 2012. Subaqueous 133e143. calving margin morphology at Mueller, Hooker and Tasman glaciers in aoraki/ Wyshnytzky, 2013. Constraining Ice Advance and Linkages to Palaeoclimate of Two Mount Cook national park, New Zealand. J. Glaciol. 58, 1037e1046. Glacial Systems in the Olympic Mountains, Washington and the Southern Alps, Rother, H., 2006. Late Pleistocene Glacial Geology of the Hope-Waiau Valley System New Zealand. MSc thesis. Utah State University. in North Canterbury, New Zealand. PhD Thesis. University of Canterbury.